正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

Today, we’re excited to launch the Datology Curation Studio, customizable frontier data curation for every team training its own models. Data quality is the ultimate compute multiplier, and Curation Studio makes it easy for everyone.

DatologyAI

3,671 subscribers

4,265,085 次观看 • 2 天前 •via X (Twitter)

新闻政治教育科学技术

Anya Rossi• Live Now

Private livecam show

39 条评论

DatologyAI 的头像

DatologyAI2 天前

We built the strongest baseline we could by combining the best publicly curated datasets. Curated data led to meaningfully better models, with the baseline needing 6× the compute to match the same quality. That’s the Datology compute multiplier in action.

DatologyAI 的头像

DatologyAI2 天前

We used the same 39 source datasets—including Nemotron, FineWeb, and The Stack v3—and identical pretraining and post-training recipes to make the most apples-to-apples comparison possible. The only change: data curation. Explore the setup and results:

DatologyAI 的头像

DatologyAI2 天前

The gains persisted after post-training: our 30B-A3B model scored 46.8% vs. 37.7% for the baseline, averaged across benchmarks. A curated 12B-A1.4B model also beat the 30B-A3B baseline despite using roughly one-fifth of the pretraining compute.

DatologyAI 的头像

DatologyAI2 天前

Curation Studio puts this process in your hands: Clean → Calibrate → Create → Compose. Bring your data, optimize for your tasks, and keep your data inside your VPC. See how it works:

DatologyAI 的头像

DatologyAI2 天前

If you want to own your own intelligence, book a demo with our team, or join us live on October 27th to see Curation Studio in action and chat with the team behind the research. Save your seat here:

Vincent Weisser 的头像

Vincent Weisser2 天前

huge congrats Ari and team!!

stochasm 的头像

stochasm2 天前

congrats on the launch!!

Felicis 的头像

Felicis2 天前

🧡🚀we love launch days!!

Astasia Myers 的头像

Astasia Myers2 天前

So cool to see this out!

Leo 的头像

Leo2 天前

this is nuts, congrats on the launch!!

z 的头像

z2 天前

same data, same recipe, just better curation and it's 6x?? y'all cooked fr

Yağız Çalık 的头像

Yağız Çalık2 天前

exciting, congrats!!

ari. 的头像

ari.2 天前

so basically it’s not about how big the model is, it’s about what you feed it. kinda makes sense when you think about it

Alba 的头像

Alba2 天前

Curious, how much does this change when you use a completely different base model?

maria 的头像

maria2 天前

everyone's buying more gpus and y'all just fixed the data lol, love this

max 的头像

max2 天前

does it work for fine-tuning too, or just pretraining?

Madeline 💌💐 的头像

Madeline 💌💐2 天前

What does “customizable” mean if you do not know what good data looks like?

Tiffany Luck 的头像

Tiffany Luck2 天前

Big day for 'own your intelligence' -- congrats!!

Rodev 的头像

Rodev2 天前

everyone wants a bigger model until they see the compute bill

rewind 的头像

rewind2 天前

Congrats team

Ryoru 𒌐 的头像

Ryoru 𒌐2 天前

Does this mean a lot of models could perform better without getting any bigger?

rachit 的头像

rachit2 天前

does the 6x hold for domain-specific data too or mostly general web?

jihyun 的头像

jihyun2 天前

Congrats @datologyai team! This is the future 🙌

Vivi 的头像

Vivi2 天前

6x from curation alone is wild

haytam 的头像

haytam2 天前

Would this help a class project, or only a team already training models?

Pablo 的头像

Pablo2 天前

Great launch guys🔥🙏🏻

nerea 的头像

nerea2 天前

love that the data stays in your own vpc, that's the part most teams actually care about

Alegna 的头像

Alegna2 天前

you guys COOKED

Kevin Moreira 的头像

Kevin Moreira2 天前

Love the focus on data quality here

alex 的头像

alex2 天前

the curated vs baseline chart is so clean, that gap says everything

lars 的头像

lars2 天前

can you plug in your own quality filters?

said 的头像

said2 天前

about time someone did this congrats guys!

Leila 👩🏻‍💻 的头像

Leila 👩🏻‍💻2 天前

Can a small team use this, or do you need a company setup?

carpe 的头像

carpe2 天前

did the baseline really need six times the compute to match that quality?

Ulises 的头像

Ulises2 天前

This is a major leap in quality.

liam. 的头像

liam.2 天前

good models start with good data, couldn't agree more

robin. 的头像

robin.2 天前

curation finally treated as a product and not a preprocessing script

Marco 的头像

Marco2 天前

congrats guys w

Stefan Friedman 的头像

Stefan Friedman2 天前

is there a minimum dataset size where this starts making sense? asking for a small team

相关视频

Data curation is crucial for post-training recipes. But how do we curate? Curation is usually manual & tedious. And, it's hard to tell if a strategy in the data will be reliable! We introduce an automatic way to curate, informed by the robot's policy learning.

Data curation is crucial for post-training recipes. But how do we curate? Curation is usually manual & tedious. And, it's hard to tell if a strategy in the data will be reliable! We introduce an automatic way to curate, informed by the robot's policy learning.

Chelsea Finn

19,910 次观看 • 1 年前

What is $REPPO? Reppo uses prediction markets to power AI data curation. Publishers submit data and domain experts stake $REPPO to assess quality within Datanets, programmable on-chain markets for any AI use case.

What is $REPPO? Reppo uses prediction markets to power AI data curation. Publishers submit data and domain experts stake $REPPO to assess quality within Datanets, programmable on-chain markets for any AI use case.

Kraken Listings

22,135 次观看 • 5 个月前

Data is the oil of modern biotech. New tools allow us to model living systems too complex for unaided human cognition. Latch is releasing a new capabilities in data curation + delivery: - 30M observational cell atlas across 150 diseases, 200 tissues, 27 technologies - a partnership with two expert labeling teams - an agentic, human-in-the-loop framework for mass scRNA-seq curation

Data is the oil of modern biotech. New tools allow us to model living systems too complex for unaided human cognition. Latch is releasing a new capabilities in data curation + delivery: - 30M observational cell atlas across 150 diseases, 200 tissues, 27 technologies - a partnership with two expert labeling teams - an agentic, human-in-the-loop framework for mass scRNA-seq curation

Kenny Workman

10,136 次观看 • 1 年前

Your community constitutes the very essence of your project and its ultimate triumph. Curation is a powerful tool for community building. With CONNECT it’s easier than ever.

Your community constitutes the very essence of your project and its ultimate triumph. Curation is a powerful tool for community building. With CONNECT it’s easier than ever.

Mindfølk+

23,467 次观看 • 3 年前

I'm excited to announce my project is an interactive art installation, "The Algorithmic Art Studio." It seeks to demonstrate the link between aesthetics, curation, generative art, and plotters. The installation will be at SIGGRAPH 2023 in LA from August 6th-10th. Details ⤵

I'm excited to announce my project is an interactive art installation, "The Algorithmic Art Studio." It seeks to demonstrate the link between aesthetics, curation, generative art, and plotters. The installation will be at SIGGRAPH 2023 in LA from August 6th-10th. Details ⤵

James Merrill

44,832 次观看 • 3 年前

Excited today to share with you the next step in the evolution of GM World Our new website features exceptional up-and-coming artists and offers more context on our curation, including the first episode of our new video series: From The Studio

Excited today to share with you the next step in the evolution of GM World Our new website features exceptional up-and-coming artists and offers more context on our curation, including the first episode of our new video series: From The Studio

BΞN

10,139 次观看 • 3 年前

Today I’m excited to introduce micro1 Intelligence, the world’s most advanced platform for training frontier AI models. Achieving AGI is bottlenecked by one main thing: high-quality data. Data based on real-world environments that capture human expert workflows, complex decision-making, and reward signals models need to learn. With micro1 Intelligence, frontier labs can train on RL environments across every subject matter all in one place.

Today I’m excited to introduce micro1 Intelligence, the world’s most advanced platform for training frontier AI models. Achieving AGI is bottlenecked by one main thing: high-quality data. Data based on real-world environments that capture human expert workflows, complex decision-making, and reward signals models need to learn. With micro1 Intelligence, frontier labs can train on RL environments across every subject matter all in one place.

Ali Ansari

333,257 次观看 • 10 个月前

Today, we're introducing AC2, the Applied Compute Agent Cloud, to enable every team to train, serve, and improve their own frontier models.

Today, we're introducing AC2, the Applied Compute Agent Cloud, to enable every team to train, serve, and improve their own frontier models.

Applied Compute

206,702 次观看 • 1 个月前

Open-ended coding training data may no longer be the bottleneck: AI can scale open-ended tasks—and even outperform human-expert curation. FrontierCS team is releasing FrontierSmith: a system for synthesizing open-ended coding problems at scale. Starting from closed-ended coding tasks, FrontierSmith mutates, filters, and builds runnable optimization environments for long-horizon coding agents. In our experiments, FrontierSmith data trains stronger models than human-curated open-ended data on FrontierCS and ALE-bench. Blog: Paper: Code: Model:

Open-ended coding training data may no longer be the bottleneck: AI can scale open-ended tasks—and even outperform human-expert curation. FrontierCS team is releasing FrontierSmith: a system for synthesizing open-ended coding problems at scale. Starting from closed-ended coding tasks, FrontierSmith mutates, filters, and builds runnable optimization environments for long-horizon coding agents. In our experiments, FrontierSmith data trains stronger models than human-curated open-ended data on FrontierCS and ALE-bench. Blog: Paper: Code: Model:

Qiuyang Mang

122,667 次观看 • 4 个月前

Much of the AI narrative revolves around architecture research, compute scaling & buying human data. Very few understand the deep technical discipline of "data research," which has driven many of the gains in model capabilities. Vivek Gopalan sits down with Mahesh Sathiamoorthy @ COLM2026 & Alex Dimakis, founders of Bespoke Labs, a company defining the category of data research. (00:00) Introducing Mahesh and Alex (03:40) Teaching a 1000 person ML class at Berkeley / changes in CS education in the age of AI (06:10) Open source data for frontier model training, OpenThoughts & training models for data curation (12:00) Why multiple-answer yields more performant models, benefits of keeping unverifiable examples in datasets (17:20) RL scaling laws / introduction to Bespoke (19:45) "Environment-driven Development" (28:35) New capabilities in frontier models & coding agents (34:30) Market outlook for RL environments (42:30) Exciting recent research and how we arrived at today's state of RL (50:00) Bespoke's Series A

Much of the AI narrative revolves around architecture research, compute scaling & buying human data. Very few understand the deep technical discipline of "data research," which has driven many of the gains in model capabilities. Vivek Gopalan sits down with Mahesh Sathiamoorthy @ COLM2026 & Alex Dimakis, founders of Bespoke Labs, a company defining the category of data research. (00:00) Introducing Mahesh and Alex (03:40) Teaching a 1000 person ML class at Berkeley / changes in CS education in the age of AI (06:10) Open source data for frontier model training, OpenThoughts & training models for data curation (12:00) Why multiple-answer yields more performant models, benefits of keeping unverifiable examples in datasets (17:20) RL scaling laws / introduction to Bespoke (19:45) "Environment-driven Development" (28:35) New capabilities in frontier models & coding agents (34:30) Market outlook for RL environments (42:30) Exciting recent research and how we arrived at today's state of RL (50:00) Bespoke's Series A

8VC

48,451 次观看 • 3 个月前

Figure just stacked the two halves of the Physical AI bottleneck a week apart. First Index: a creator network collecting real-world task video at roughly half an hour of uploads every second, across homes and workplaces in 100+ countries. That is the long-tail data you do not get from a lab teleop setup. Then Nscale: $3.5B of compute to start, intent past $6B, up to 100,000 NVIDIA Vera Rubin GPUs. Brett Adcock said they are largely constrained by data and compute. Index is the data bet. Nscale is the compute bet. Figure's own writeup says the quiet part: data alone cannot solve this. My take: Compute is a prerequisite. Collecting more data is critical. But finding the 1% of data that actually teaches the model is what compounds. Search and curation, not just capture. That's the problem we are focused on solving at Foxglove.

Figure just stacked the two halves of the Physical AI bottleneck a week apart. First Index: a creator network collecting real-world task video at roughly half an hour of uploads every second, across homes and workplaces in 100+ countries. That is the long-tail data you do not get from a lab teleop setup. Then Nscale: $3.5B of compute to start, intent past $6B, up to 100,000 NVIDIA Vera Rubin GPUs. Brett Adcock said they are largely constrained by data and compute. Index is the data bet. Nscale is the compute bet. Figure's own writeup says the quiet part: data alone cannot solve this. My take: Compute is a prerequisite. Collecting more data is critical. But finding the 1% of data that actually teaches the model is what compounds. Search and curation, not just capture. That's the problem we are focused on solving at Foxglove.

Adrian Macneil

104,588 次观看 • 1 个月前

Scale alone is not enough for AI data. Quality and complexity are equally critical. Excited to support all of these for LLM developers with Snorkel AI Data-as-a-Service, and to share our new leaderboard! — Our decade-plus of research and work in AI data has a simple point: scale alone is not enough. AI success is all about the quality, complexity, and distribution of data—in addition to volume. We’re excited to be powering leading LLM developers with Snorkel AI Expert Data-as-a-Service, our white glove service for custom, expert-level AI datasets—and to now preview some of what we’re building via our new Expert Data Leaderboard (🔗 in 🧵) + upcoming OSS dataset releases! Snorkel Expert Data-as-a-Service is built to meet the rapidly evolving data needs of the agentic AI world—where success is built on the quality, complexity, and distribution of datasets, in addition to size and scale. This kind of high-quality, frontier AI data can only come from a union of technology and human expertise. With Snorkel Expert Data-as-a-Service, we’re powering frontier LLM developers across agentic, expert knowledge, reasoning, coding, multi-modal, and other task types via the combination of these two key components: - (1) The Snorkel Expert Network: A global team of subject matter experts focused wholly on specialized knowledge–spanning thousands of topics in STEM/academic, vertical/professional, and consumer/lifestyle domains. - (2) Snorkel AI Data Development Platform: Our unique programmatic data curation and quality control platform, accelerating and improving expert authoring and review through principled techniques developed over the last decade of R&D. Now: we’re incredibly excited to showcase some of the power of Snorkel Expert Data-as-a-Service via the new Snorkel Leaderboard—putting frontier models to the test in complex, agentic, and reasoning settings inspired by real industry scenarios (not esoteric puzzles)! We’ll be releasing new leaderboards and accompanying expert-verified open source datasets (coming soon!) regularly. To start, we’re sharing three initial ones in preview: - SnorkelFinance: Q&A over financial documents requiring agentic tool-calling and reasoning - SnorkelUnderwrite: Agentic insurance tasks requiring industry-specific reasoning and tool use - SnorkelSequences: Mathematical tasks requiring compositional multi-step reasoning

Scale alone is not enough for AI data. Quality and complexity are equally critical. Excited to support all of these for LLM developers with Snorkel AI Data-as-a-Service, and to share our new leaderboard! — Our decade-plus of research and work in AI data has a simple point: scale alone is not enough. AI success is all about the quality, complexity, and distribution of data—in addition to volume. We’re excited to be powering leading LLM developers with Snorkel AI Expert Data-as-a-Service, our white glove service for custom, expert-level AI datasets—and to now preview some of what we’re building via our new Expert Data Leaderboard (🔗 in 🧵) + upcoming OSS dataset releases! Snorkel Expert Data-as-a-Service is built to meet the rapidly evolving data needs of the agentic AI world—where success is built on the quality, complexity, and distribution of datasets, in addition to size and scale. This kind of high-quality, frontier AI data can only come from a union of technology and human expertise. With Snorkel Expert Data-as-a-Service, we’re powering frontier LLM developers across agentic, expert knowledge, reasoning, coding, multi-modal, and other task types via the combination of these two key components: - (1) The Snorkel Expert Network: A global team of subject matter experts focused wholly on specialized knowledge–spanning thousands of topics in STEM/academic, vertical/professional, and consumer/lifestyle domains. - (2) Snorkel AI Data Development Platform: Our unique programmatic data curation and quality control platform, accelerating and improving expert authoring and review through principled techniques developed over the last decade of R&D. Now: we’re incredibly excited to showcase some of the power of Snorkel Expert Data-as-a-Service via the new Snorkel Leaderboard—putting frontier models to the test in complex, agentic, and reasoning settings inspired by real industry scenarios (not esoteric puzzles)! We’ll be releasing new leaderboards and accompanying expert-verified open source datasets (coming soon!) regularly. To start, we’re sharing three initial ones in preview: - SnorkelFinance: Q&A over financial documents requiring agentic tool-calling and reasoning - SnorkelUnderwrite: Agentic insurance tasks requiring industry-specific reasoning and tool use - SnorkelSequences: Mathematical tasks requiring compositional multi-step reasoning

Alex Ratner

495,851 次观看 • 1 年前

.L3Harris says Palantir helped them beat frontier AI models in less than 2 days, at 95% lower cost: "When we fine-tuned open source models trained on our own data, we were able to outperform the frontier models in less than 48 hours." "The cost of our fine-tuned open source model was 95% lower than the frontier models we were using." "AI is a commodity. It's all about the data. We view our data as a corporate asset. It's our unique hard-earned knowledge." "We believe American defense companies should not be a vassal for frontier AI labs, handing over our data and institutional knowledge, hoping to rent back the intelligence it creates." " We own the model, we own the compute, we own the advantage."

.L3Harris says Palantir helped them beat frontier AI models in less than 2 days, at 95% lower cost: "When we fine-tuned open source models trained on our own data, we were able to outperform the frontier models in less than 48 hours." "The cost of our fine-tuned open source model was 95% lower than the frontier models we were using." "AI is a commodity. It's all about the data. We view our data as a corporate asset. It's our unique hard-earned knowledge." "We believe American defense companies should not be a vassal for frontier AI labs, handing over our data and institutional knowledge, hoping to rent back the intelligence it creates." " We own the model, we own the compute, we own the advantage."

Jawwwn

1,106,105 次观看 • 1 个月前

No one else can do this. While every other CV platform is still selling another fine-tuning dashboard We’re about to let anyone order actual frontier lab work in a few clicks. The kind only Bittensor miners can deliver. Model Foundry is coming to Score Studio. Zero compute? Zero data? Zero ML team? Doesn’t matter. You describe the hard vision problem, our miners build the production model, you pay inside the Studio. That spend goes straight into SN44 buyback + burn. Money never leaves the subnet. No gatekeepers. No bottlenecks. Just open frontier vision for everyone. Check out the Model Foundry flow 👇

No one else can do this. While every other CV platform is still selling another fine-tuning dashboard We’re about to let anyone order actual frontier lab work in a few clicks. The kind only Bittensor miners can deliver. Model Foundry is coming to Score Studio. Zero compute? Zero data? Zero ML team? Doesn’t matter. You describe the hard vision problem, our miners build the production model, you pay inside the Studio. That spend goes straight into SN44 buyback + burn. Money never leaves the subnet. No gatekeepers. No bottlenecks. Just open frontier vision for everyone. Check out the Model Foundry flow 👇

Score

24,327 次观看 • 1 个月前

Traverse (Traverse (YC W26)) is a data research lab solving the hardest problem in AI training: subjective, taste-dependent work. They generate long-horizon training data for frontier models by observing and capturing what specialists actually do on the job. Congrats on the launch, Lance Yan and Zachary Yu!

Traverse (Traverse (YC W26)) is a data research lab solving the hardest problem in AI training: subjective, taste-dependent work. They generate long-horizon training data for frontier models by observing and capturing what specialists actually do on the job. Congrats on the launch, Lance Yan and Zachary Yu!

Y Combinator

78,149 次观看 • 7 个月前

Developers are coming to Avalanche for AI; it offers customizable compute and data structures, plus a robust ecosystem across finance, gaming, payments, and more. Listen to KITE AI talk about how this synergy is powering new AI agents and models for every sector. For more: 👇

Developers are coming to Avalanche for AI; it offers customizable compute and data structures, plus a robust ecosystem across finance, gaming, payments, and more. Listen to KITE AI talk about how this synergy is powering new AI agents and models for every sector. For more: 👇

Avalanche🔺

22,981 次观看 • 1 年前

Excited to launch FactIQ today! 🚀 We just indexed 7.4M+ official US data series to build the ultimate economic research agent. Visualize trends instantly. Verify every source. Export charts for your reports. Free for the next week - try it out at factiq[dot]com!

Excited to launch FactIQ today! 🚀 We just indexed 7.4M+ official US data series to build the ultimate economic research agent. Visualize trends instantly. Verify every source. Export charts for your reports. Free for the next week - try it out at factiq[dot]com!

FactIQ

26,356 次观看 • 10 个月前

Argus: an open-source robotics data annotation and quality pipeline. With new frontier VLMs like GPT-6 Astra, we can generate rich, high-quality annotations for robotics data for both training and dataset analysis. Argus delivers detailed, timestamp-level annotations while catching issues like mislabeled instructions, sped-up recordings, swapped camera streams, and unflagged operator mistakes. To support better data quality for the robotics community, we’re open-sourcing Argus for anyone to use 🧵

Argus: an open-source robotics data annotation and quality pipeline. With new frontier VLMs like GPT-6 Astra, we can generate rich, high-quality annotations for robotics data for both training and dataset analysis. Argus delivers detailed, timestamp-level annotations while catching issues like mislabeled instructions, sped-up recordings, swapped camera streams, and unflagged operator mistakes. To support better data quality for the robotics community, we’re open-sourcing Argus for anyone to use 🧵

calix

84,347 次观看 • 10 天前

Why is my large purposefully curated OFFLINE data so valuable? Listen to someone who build the common crawl. My data is unchanged with receipts and most of it has not been seen for decades. Online training data is the lowest quality data.

Why is my large purposefully curated OFFLINE data so valuable? Listen to someone who build the common crawl. My data is unchanged with receipts and most of it has not been seen for decades. Online training data is the lowest quality data.

Brian Roemmele

25,070 次观看 • 7 个月前

I DID IT! I now have 512 streams of live talk radio from around the world converted in real-time for AI training. I started using HAM, CB, Shortwave and independent talk radio about 15 years ago for future AI training. The largest curation in the world!

I DID IT! I now have 512 streams of live talk radio from around the world converted in real-time for AI training. I started using HAM, CB, Shortwave and independent talk radio about 15 years ago for future AI training. The largest curation in the world!

Brian Roemmele

88,522 次观看 • 11 个月前